Thank you for the report. So you have a minimal repro to reproduce the issue you are seeing?
On Friday, April 9, 2021, 谢旗旺 <1415850...@qq.com> wrote: > When I use "pd.read_parquet(path,engine='pyarrow’)” to load my > dataset,the loaded dataset appears where one piece of data is copied into > two，but fastparquet will not. >
